Although many pronunciation instructions in Japan focus on segmental aspects, most of the pronunciation difficulties arise from the difference of perceptual prosodic systems. This paper analyses the pronunciation problems observed in typical Japanese learners of English, and examines what underlies such problematic performance. It gives two kinds of prosodic discrepancy as the major causes. The first cause is the difference in basic rhythm-counting units: mora in Japanese and syllable in English. It leads to various surface problems such as inability to link words, insertion of a vocalic segment after every non-moraic consonant, abrupt pitch change in the production of long vowels, and other prosodic troubles as well as segmental difficulties. Our second argument draws on the difference of how prominence is realised on words. Japanese is a pitch-accent language and it differs from stress-accent English in that the former only employs pitch to show prominent elements whereas the latter combines pitch, loudness, and duration. Another difference in terms of this is that English has an alternating stress pattern, while Japanese prominent elements appear in clusters. These cause the learners' monotonous performance, confusion of long and short vowels, inability to distinguish strong and weak syllables, and other problems concerning sentence rhythm. We also try to link the causes and the problems by giving explanations from phonetic and phonological theory and practice.